US8254718B2ActiveUtilityPatentIndex 72
Multi-channel edge-aware chrominance noise reduction
Est. expiryMay 15, 2028(~1.9 yrs left)· nominal 20-yr term from priority
G06V 10/30G06T 5/50G06T 2207/20028G06T 2207/20032G06T 5/20G06T 2207/20192G06T 2207/10024G06T 5/70
72
PatentIndex Score
20
Cited by
25
References
22
Claims
Abstract
Noise in an image is reduced in a manner that takes into account edge information in one or more channels of the image. A first image is received that is formatted according to a red-green-blue (RGB) color model. The first image is converted from the RGB color model to a second color model that includes at least a luminance channel, a first chrominance channel, and a second chrominance channel that are representative of the first image. The first and second chrominance channels are each denoised in a manner that accounts at least for edge information in the luminance channel, and may also include edge information from other channels in a manner that accounts for per-channel noise characteristics. The luminance channel and denoised first and second chrominance channels are converted to a second image formatted according to the RGB color model that is a noise-reduced version of the first image.
Claims
exact text as granted — not AI-modified1. A method for reducing noise in an image, comprising:
receiving a first image as a first set of data formatted according to a red-green-blue (RGB) color model;
converting the first set of data from the RGB color model to a second color model that includes at least a luminance channel, a first chrominance channel, and a second chrominance channel that are representative of the first image;
denoising the first chrominance channel using modified joint bilateral filtering in a manner that accounts at least for edge information in the luminance channel by calculating a denoised chrominance value for each pixel of the first chrominance channel according to
J
s
C
1
=
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
I
p
C
1
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
,
where
I C1 is the first chrominance channel,
Ω is a region of pixels of the first chrominance channel that includes a pixel s,
p is a pixel located in the region of pixels Ω,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I p L is a luminance value in the luminance channel for the pixel p,
I s L is a luminance value in the luminance channel for the pixel s,
J C1 is the denoised first chrominance channel image,
J s C1 is the calculated denoised chrominance value of the pixel s,
∥p-s∥ is an Euclidean distance between the pixel p and the pixel s,
σ h is a spatial falloff parameter,
σ i L is an intensity falloff parameter for the luminance channel, and
g( ) is a Gaussian distribution;
denoising the second chrominance channel in a manner that accounts at least for edge information in the luminance channel; and
converting the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second set of data formatted according to the RGB color model that is representative of a second image.
2. The method of claim 1 , further comprising:
processing the luminance channel; and
wherein said converting the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second set of data formatted according to the RGB color model that is representative of a second image comprises
converting the processed luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to the second set of data formatted according to the RGB color model.
3. The method of claim 1 , further comprising:
median filtering the luminance channel; and
wherein said denoising the first chrominance channel using modified joint bilateral filtering in a manner that accounts at least for edge information in the luminance channel comprises:
denoising the first chrominance channel modified joint bilateral filtering in a manner that accounts for edge information in the filtered luminance channel; and
said denoising the second chrominance channel in a manner that accounts at least for edge information in the luminance channel comprises:
denoising the second chrominance channel in a manner that accounts for edge information in the filtered luminance channel.
4. The method of claim 1 , further comprising:
median filtering the luminance channel; and
wherein said denoising the first chrominance channel using modified joint bilateral filtering in a manner that accounts at least for edge information in the luminance channel comprises:
denoising the first chrominance channel modified joint bilateral filtering in a manner that accounts for edge information in the filtered luminance channel; and
said denoising the second chrominance channel in a manner that accounts at least for edge information in the luminance channel comprises:
denoising the second chrominance channel in a manner that accounts for edge information in the filtered luminance channel.
5. A method for reducing noise in an image, comprising:
converting a first image formatted according to a first color model from a chrominance-specific color model to a second color model that includes at least a luminance channel, a first chrominance channel, and a second chrominance channel that are representative of the first image;
denoising the first chrominance channel using modified dual bilateral filtering in a manner that accounts for edge information in the luminance channel and edge information in either the first chrominance channel or the second chrominance channel by calculating a denoised chrominance value for each pixel of the first chrominance channel according to
J
s
C
1
=
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
g
(
I
p
C
1
-
I
s
C
1
,
σ
i
C
1
)
I
p
C
1
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
g
(
I
p
C
1
-
I
s
C
1
,
σ
i
C
1
)
,
where
I C1 is the first chrominance channel,
Ω is a region of pixels of the first chrominance channel that includes a pixel s,
p is a pixel located in the region of pixels Ω,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I p L is a luminance value in the luminance channel for the pixel p,
I s L is a luminance value in the luminance channel for the pixel s,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I s C1 is a chrominance value in the first chrominance channel for the pixel s,
J C1 is the denoised first chrominance channel image,
J s C1 is the calculated denoised chrominance value of the pixel s,
∥p-s∥ is an Euclidean distance between the pixel p and the pixel s,
σ h is a spatial falloff parameter,
σ i L is an intensity falloff parameter for the luminance channel,
σ i C1 is an intensity falloff parameter for the first chrominance channel, and
g( ) is a Gaussian distribution;
denoising the second chrominance channel in a manner that accounts for edge information in the luminance channel and edge information in either the first chrominance channel or the second chrominance channel; and
converting the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second image formatted according to the first color model.
6. The method of claim 5 , further comprising:
processing the luminance channel; and
wherein said converting the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second set of data formatted according to the first color model that is representative of a second image comprises
converting the processed luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to the second set of data formatted according to the first color model.
7. A method for reducing noise in an image, comprising:
receiving a first image as a first set of data formatted according to a red-green-blue (RGB) color model;
converting the first set of data from the RGB color model to a second color model that includes at least a luminance channel, a first chrominance channel, and a second chrominance channel that are representative of the first image;
denoising the first chrominance channel using trilateral filtering in a manner that accounts for edge information in the luminance channel, edge information in the first chrominance channel, and edge information in the second chrominance channel by calculating a denoised chrominance value for each pixel of the first chrominance channel according to
J
s
C
1
=
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
g
(
I
p
C
1
-
I
s
C
1
,
σ
i
C
1
)
g
(
I
p
C
2
-
I
s
C
2
,
σ
i
C
2
)
I
p
C
1
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
g
(
I
p
C
1
-
I
s
C
1
,
σ
i
C
1
)
g
(
I
p
C
2
-
I
s
C
2
,
σ
i
C
2
)
,
where
I C1 is the first chrominance channel,
Ω is a region of pixels of the first chrominance channel that includes a pixel s,
p is a pixel located in the region of pixels Ω,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I p L is a luminance value in the luminance channel for the pixel p,
I s L is a luminance value in the luminance channel for the pixel s,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I s C1 is a chrominance value in the first chrominance channel for the pixel s,
I p C2 is a chrominance value in the second chrominance channel for the pixel p,
I s C2 is a chrominance value in the second chrominance channel for the pixel s,
J C1 is the denoised first chrominance channel image,
J s C1 is the calculated denoised chrominance value of the pixel s,
∥p-s∥ is an Euclidean distance between the pixel p and the pixel s,
σ h is a spatial falloff parameter,
σ i L is an intensity falloff parameter for the luminance channel,
σ i C1 is an intensity falloff parameter for the first chrominance channel,
σ i C2 is an intensity falloff parameter for the second chrominance channel, and
g( ) is a Gaussian distribution;
denoising the second chrominance channel in a manner that accounts for edge information in the luminance channel, edge information in the first chrominance channel, and edge information in the second chrominance channel; and
converting the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second set of data formatted according to the RGB color model that is representative of a second image.
8. The method of claim 7 , further comprising:
processing the luminance channel; and
wherein said converting the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second set of data formatted according to the RGB color model that is representative of a second image comprises
converting the processed luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to the second set of data formatted according to the RGB color model.
9. A system for reducing noise in an image, comprising:
a first color model converter configured to receive a first image as a first set of data formatted according to a red-green-blue (RGB) color model, and to convert the first set of data from the RGB color model to a second color model that includes at least a luminance channel, a first chrominance channel, and a second chrominance channel that are representative of the first image;
a first chrominance channel filter configured to denoise the first chrominance channel using modified joint bilateral filtering in a manner that accounts at least for edge information in the luminance channel by being configured to calculate a denoised chrominance value for each pixel of the first chrominance channel according to
J
s
C
1
=
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
I
p
C
1
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
,
where
I C1 is the first chrominance channel,
Ω is a region of pixels of the first chrominance channel that includes a pixel s,
p is a pixel located in the region of pixels Ω,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I p L is a luminance value in the luminance channel for the pixel p,
I s L is a luminance value in the luminance channel for the pixel s,
J C1 is the denoised first chrominance channel image,
J s C1 is the calculated denoised chrominance value of the pixel s,
∥p-s∥ is an Euclidean distance between the pixel p and the pixel s,
σ h is a spatial falloff parameter,
σ i L is an intensity falloff parameter for the luminance channel, and
g( ) is a Gaussian distribution;
a second chrominance channel filter configured to denoise the second chrominance channel in a manner that accounts at least for edge information in the luminance channel; and
a second color model converter configured to convert the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second set of data formatted according to the RGB color model that is configured to form a second image.
10. The system of claim 9 , further comprising:
a luminance channel processor configured to process the luminance channel; and
the second color model converter being configured to convert the processed luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to the second set of data formatted according to the RGB color model to form the second image.
11. The system of claim 9 , further comprising:
a median filter configured to filter the luminance channel.
12. The system of claim 9 , wherein the first chrominance channel filter is configured to denoise the first chrominance channel in a manner that accounts for edge information in the luminance channel and edge information in the first chrominance channel; and
the second chrominance channel filter is configured to denoise the second chrominance channel in a manner that accounts for edge information in the luminance channel and edge information in the second chrominance channel.
13. The system of claim 9 , further comprising:
a median filter configured to filter the luminance channel.
14. A system for reducing noise in an image, comprising:
a first color model converter configured to receive a first image as a first set of data formatted according to a first color model from a chrominance-specific color model, and to convert the first set of data from the first color model to a second color model that includes at least a luminance channel, a first chrominance channel, and a second chrominance channel that are representative of the first image;
a first chrominance channel filter configured to denoise the first chrominance channel using modified dual bilateral filtering in a manner that accounts for edge information in the luminance channel and edge information in either the first chrominance channel or the second chrominance channel by being configured to calculate a denoised chrominance value for each pixel of the first chrominance channel according to
J
s
C
1
=
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
g
(
I
p
C
1
-
I
s
C
1
,
σ
i
C
1
)
I
p
C
1
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
g
(
I
p
C
1
-
I
s
C
1
,
σ
i
C
1
)
,
where
I C1 is the first chrominance channel,
Ω is a region of pixels of the first chrominance channel that includes a pixel s,
p is a pixel located in the region of pixels Ω,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I p L is a luminance value in the luminance channel for the pixel p,
I s L is a luminance value in the luminance channel for the pixel s,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I s C1 is a chrominance value in the first chrominance channel for the pixel s,
J C1 is the denoised first chrominance channel image,
J s C1 is the calculated denoised chrominance value of the pixel s,
∥p-s∥ is an Euclidean distance between the pixel p and the pixel s,
σ h is a spatial falloff parameter,
σ i C1 is an intensity falloff parameter for the luminance channel,
σ i C1 is an intensity falloff parameter for the first chrominance channel, and
g( ) is a Gaussian distribution;
a second chrominance channel filter configured to denoise the second chrominance channel in a manner that accounts for edge information in the luminance channel and edge information in either the first chrominance channel or the second chrominance channel; and
a second color model converter configured to convert the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second set of data formatted according to the first color model that is configured to form a second image.
15. The method of claim 14 , further comprising:
processing the luminance channel; and
wherein said converting the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second image formatted according to the first color model comprises
converting the processed luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to the second image formatted according to the first color model.
16. The method of claim 14 , further comprising:
median filtering the luminance channel; and
wherein said denoising the first chrominance channel using modified dual bilateral filtering in a manner that accounts for edge information in the luminance channel and edge information in either the first chrominance channel or the second chrominance channel comprises
denoising the first chrominance channel using modified dual bilateral filtering in a manner that accounts for edge information in the filtered luminance channel and edge information in either the first chrominance channel or the second chrominance channel; and
wherein said denoising the second chrominance channel in a manner that accounts for edge information in the luminance channel and edge information in either the first chrominance channel or the second chrominance channel comprises
denoising the second chrominance channel in a manner that accounts for edge information in the filtered luminance channel and edge information in either the first chrominance channel or the second chrominance channel.
17. The system of claim 14 , further comprising:
a luminance channel processor configured to process the luminance channel; and
the second color model converter being configured to convert the processed luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to the second set of data formatted according to the first color model to form the second image.
18. The method of claim 14 , further comprising:
median filtering the luminance channel; and
wherein said denoising the first chrominance channel using modified dual bilateral filtering in a manner that accounts for edge information in the luminance channel and edge information in either the first chrominance channel or the second chrominance channel comprises
denoising the first chrominance channel using modified dual bilateral filtering in a manner that accounts for edge information in the filtered luminance channel and edge information in either the first chrominance channel or the second chrominance channel; and
wherein said denoising the second chrominance channel in a manner that accounts for edge information in the luminance channel and edge information in either the first chrominance channel or the second chrominance channel comprises
denoising the second chrominance channel in a manner that accounts for edge information in the filtered luminance channel and edge information in either the first chrominance channel or the second chrominance channel.
19. A system for reducing noise in an image, comprising:
a first color model converter configured to receive a first image as a first set of data formatted according to a red-green-blue (RGB) color model, and to convert the first set of data from the RGB color model to a second color model that includes at least a luminance channel, a first chrominance channel, and a second chrominance channel that are representative of the first image;
a first chrominance channel filter configured to denoise the first chrominance channel using trilateral filtering in a manner that accounts for edge information in the luminance channel, edge information in the first chrominance channel, and edge information in the second chrominance channel by being configured to calculate a denoised chrominance value for each pixel of the first chrominance channel according to
J
s
C
1
=
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
g
(
I
p
C
1
-
I
s
C
1
,
σ
i
C
1
)
g
(
I
p
C
2
-
I
s
C
2
,
σ
i
C
2
)
I
p
C
1
∑
p
=
Ω
g
(
p
-
s
,
σ
h
)
g
(
I
p
L
-
I
s
L
,
σ
i
L
)
g
(
I
p
C
1
-
I
s
C
1
,
σ
i
C
1
)
g
(
I
p
C
2
-
I
s
C
2
,
σ
i
C
2
)
,
where
I C1 is the first chrominance channel,
Ω is a region of pixels of the first chrominance channel that includes a pixel s,
p is a pixel located in the region of pixels Ω,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I p L is a luminance value in the luminance channel for the pixel p,
I p L is a luminance value in the luminance channel for the pixel s,
I p C1 is a chrominance value in the first chrominance channel for the pixel p,
I s C1 is a chrominance value in the first chrominance channel for the pixel s,
I p C2 is a chrominance value in the second chrominance channel for the pixel p,
I s C2 is a chrominance value in the second chrominance channel for the pixel s,
J C1 is the denoised first chrominance channel image,
J s C1 is the calculated denoised chrominance value of the pixel s,
∥p-s∥ is an Euclidean distance between the pixel p and the pixel s,
σ h is a spatial falloff parameter,
σ i L is an intensity falloff parameter for the luminance channel,
σ i C1 is an intensity falloff parameter for the first chrominance channel,
σ i C2 is an intensity falloff parameter for the second chrominance channel, and
g( ) is a Gaussian distribution;
a second chrominance channel filter configured to denoise the second chrominance channel in a manner that accounts for edge information in the luminance channel, edge information in the first chrominance channel, and edge information in the second chrominance channel; and
a second color model converter configured to convert the luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to a second set of data formatted according to the RGB color model that is configured to form a second image.
20. The method of claim 19 , further comprising:
capturing a third image in the form of a two-dimensional array of pixels arranged in a Bayer pattern; and
interpolating the two-dimensional array of pixels arranged in a Bayer pattern to form the first image formatted according to the RGB color model.
21. The system of claim 19 , further comprising:
a luminance channel processor configured to process the luminance channel; and
the second color model converter being configured to convert the processed luminance channel, the denoised first chrominance channel, and the denoised second chrominance channel to the second set of data formatted according to the RGB color model to form the second image.
22. The method of claim 19 , further comprising:
capturing a third image in the form of a two-dimensional array of pixels arranged in a Bayer pattern; and
interpolating the two-dimensional array of pixels arranged in a Bayer pattern to form the first image formatted according to the RGB color model.Cited by (0)
No later patents cite this yet.
References (0)
No backward citations on record.